一种基于人眼视觉系统特性的粒子滤波跟踪方法  

A particle filter tracking method based on human visual system characters

作　　者：卢宇[1,2] 祝豪[1,2] 吴钦章[1] 

机构地区：[1]中国科学院光电技术研究所,成都610209 [2]中国科学院研究生院,北京100039

出　　处：《光学技术》2009年第2期213-216,220,共5页Optical Technique

摘　　要：通过模仿人眼跟踪运动目标的特点,对粒子滤波框架进行改造,提出了一种基于人眼视觉系统特性的粒子滤波跟踪算法。该算法采用多特征融合,将颜色特征与运动面积特征相结合,对运动区域比较敏感;在状态更新阶段,对跟踪粒子进行进一步筛选,自适应地选择性能良好的粒子进行结果判决,保障了跟踪结果的准确性。实验表明,该方法不仅可行,而且具有良好的跟踪效果。On the basis of simulating the human eye＇s characters, in the process of the object tracking and modifying the particle filter framework, a particle filter tracking method based on human visual system characters is proposed. This method adopts the multi-cue fusion and utilizes the combination between the grey histogram and the motion region area as the criteria of the similarity of each particle, which improves the sensitivity about the region where the motion takes place. Particles are filtrated in the update stage and only the ones with comparative larger weight are picked out to judge the final tracking result, which guarantees the tracking precision. Experimental results show this tracking method is not only feasible but also with good performance.

关 键 词：目标跟踪 人眼视觉系统(Human VISUAL system HVS) 粒子滤波 多特征融合 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]